AIML@TU club logoAIML@TU
Graduate Research Showcase

RESEARCHMIXER.

Grad students from every discipline — come learn about and discuss AI/ML research happening at TU.

  • AI/ML
  • CS
  • PE
  • ME
  • ECE
  • Cyber
  • Bio
  • ChemE
  • Physics
  • + yours
You pick the format.Vote on the RSVP
A

Showcase

2–3 students present, then Q&A.

B

Mixer

Icebreaker — everyone shares their research.

C

Your idea

Suggest your own format.

WED OCT 14 · 12–12:50 PM
Graves Commons, Rayzor Hall · LUNCH SERVED
Undergrads & faculty welcome · @tu.aiml.club
QR code that opens the AIML@TU meeting attendance sign-in form
Sign in
Who we are

AIML@TU

TU's student-led community for AI and machine learning
  • Research, innovation and discussion, across every field of study
  • Founded Fall 2024 · SGA Best New Student Organization, 2025
  • All majors, freshmen to PhD candidates. No experience required.
What we do

Learn, build, connect

  • Lunch & Learns every month: research talks, workshops and guests
  • AI Speaker Series with Tulsa Innovation Labs
  • Multi-Club Mashup, the campus-wide club showcase we host
  • Hurricane Hackathon, TU's first hackathon, May 2026
AIML@TU Fall 2026 events: Tier Lists Sept 4, AI Speaker Series with Tyrance Billingsley II Sept 16, Multi-Club Mashup IV Sept 29, Graduate Research Showcase Oct 14
A lightsaber demonstration on stage at Multi-Club Mashup IV, with a slide of sword-fighting tips on the screen behind
Sept 29 · Multi-Club Mashup IV Clubs from across campus, one Great Hall Spotlights, Ultra Jeopardy and the AI Slop-etition
Coming up

What's next

  • More AI Speaker Series talks this fall
  • Hurricane Hackathon II in Spring 2027. Dates to come.
  • Want to help plan them? Talk to an officer.
Hurricane Hackathon participants waving at the camera in Keplinger Hall
Today

A bit of both

A
Showcase
Three research talks, with time for questions
B
Mixer
Warm-up, 10 minutes
Roundtable, 10 to 15 minutes
Research talk

Saving the Planet, One Token at a Time

Zach Gray
Open the talk
Dapeng Zhang, slide 1. Title slide: Inference-Time Self-Improvement in Language Models, Dapeng Zhang, PhD student, School of Cyber Studies, The University of Tulsa
Dapeng Zhang, slide 2. Context: mainstream LLM improvement techniques, split into those that change weights and those that keep weights frozen
Dapeng Zhang, slide 3. Memory-based continual adaptation: Dynamic Cheatsheet, with solution generation and memory curation
Dapeng Zhang, slide 4. Memory-based continual adaptation: Dynamic Cheatsheet, continued
Dapeng Zhang, slide 5. Memory-based continual adaptation: Agentic Context Engineering (ACE), with generator, reflector and curator
Dapeng Zhang, slide 6. Memory-based continual adaptation: Agentic Context Engineering (ACE), continued
Dapeng Zhang, slide 7. Our proposed framework: Learn to Remember (LeRe), with planner, solver, curator and a memory bank
Dapeng Zhang, slide 8. Our proposed framework: Learn to Remember (LeRe), accuracy results across three models
Dapeng Zhang, slide 9. Thanks for listening! Any questions?
Nouman Ahmed, slide 1. Title slide: Talking to a Recommender. Using an LLM as a semantic translator for a collaborative-filtering movie recommender. Nouman Ahmed, advisor Dr. Sandip Sen, The University of Tulsa, 2025
Nouman Ahmed, slide 2. The problem: recommenders speak structure, people speak language
Nouman Ahmed, slide 3. Research objective: use the LLM as a semantic translator, in four steps: listen, translate, retrieve, explain
Nouman Ahmed, slide 4. Related work: three roles LLMs play in recommendation
Nouman Ahmed, slide 5. System design: how a request flows from the user through an LLM parser, ID mapping, the recommender and an LLM responder
Nouman Ahmed, slide 6. Experimental setup: MovieLens 100K, BPR and CDAE recommenders, gpt-4o-mini
Nouman Ahmed, slide 7. What happened: the user asked for action movies and the system heard something else
Nouman Ahmed, slide 8. Findings: both halves of the system fall short
Nouman Ahmed, slide 9. Future work: tighten the link between intent and retrieval
Nouman Ahmed, slide 10. Takeaway: the LLM can hold the conversation, the recommender has to hear it. Thank you. Questions?
Warm-up · 10 minutes

Find someone you haven't met

  • What are you excited about?
  • What are you scared about?
Roundtable · 10 to 15 minutes

Around the room

  • Who are you, and what degree are you pursuing?
  • What's your research area?
  • What else are you interested in?
Looking for a collaborator? Say so.
Graduate Research Showcase

Keep talking

Swap contacts. Find a collaborator.
Want to present next time? Talk to an officer.
QR code that opens the AIML@TU meeting attendance sign-in form
Sign in
QR code that opens the AIML@TU club website
Club website